The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses system commands to gather local diagnostic information, such as the operating system name and version, to include in the bug report.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied descriptions and local file content to populate issue templates, which is a common surface for indirect prompt injection.
Ingestion points: User input describing a bug or feature, and the contents of the plugin.json configuration file located in the plugin root.
Boundary markers: The skill is strictly instructed to display the complete proposed payload in a fenced text block for user inspection before any data leaves the local environment.
Capability inventory: The skill uses Bash for system diagnostics and Read for accessing the version manifest.
Sanitization: The instructions include a comprehensive list of data to be scrubbed, including PII (names, addresses, emails), environment variables, secrets, tokens, and proprietary document content.

studio-feedback